Will the next Google Gemini Pro model added to the Arena Leaderboard debut at a score of at least 1490?

- [2026-02-19/20] Google released Gemini 3.1 Pro (preview); on the Arena leaderboard it debuted tied #1 in Text (scoring 1500), 4 pts from Opus 4.6, and top 3 in the Arena Expert Leaderboard (scoring 1538) . This debut predates the market's May 18, 2026 open, so it is not the "next" model this market resolves on. - [2026-05-19] At Google I/O, Google announced Gemini 3.5 Pro as its next flagship, promising a 2-million-token context window and a built-in Deep Think mode; it is Google's announced next flagship model, unveiled at Google I/O on 19 May 2026, with a 2-million-token context window and a built-in Deep Think reasoning mode . - [2026-08-23] Gemini 3.5 Pro remained unshipped: it was still unreleased as of 23 August 2026, more than three months after Google announced it at I/O on 19 May, having now missed three targets (June, mid-July, early August), with no model ID, no pricing and no launch date . - [Late Aug 2026] Reports indicate Gemini 3.5 Pro was cancelled: the Wall Street Journal reported in late August that Gemini 3.5 Pro had been scrapped because internal candidates "weren't sufficiently better than the Flash series" . Another source similarly says Gemini 3.5 Pro Shelved: Google quietly canceled Gemini 3.5 Pro to channel all compute resources directly into Gemini 4 . - [2026-07-21] Google publicly confirmed the next generational model is in training: on July 21, 2026, in the post announcing Gemini 3.6 Flash, Google DeepMind wrote that it had "started our most ambitious pre-training run yet, for Gemini 4," and Sundar Pichai read almost the same words to analysts on Alphabet's second-quarter call the next day . - [2026-09-02] Google shipped Gemini 3.8 Flash (a Flash, not Pro, model), which does not qualify for this market: per Wikipedia, the stable release listed is 3.1 Pro, 3 Deep Think, 3.8 Flash, 3.5 Flash-Lite / September 2, 2026 (3.8 Flash) . - [2026-09-13] Current lmarena.ai Text Arena Overall leaderboard (fetched directly) shows no new Gemini Pro model since 3.1 Pro; top entries include claude-fable-5-high at 1506, claude-opus-4-6-high at 1505, with gemini-3.1-pro-preview at 1487 and gemini-3-pro at 1485 further down; the highest Gemini entry overall is gemini-3.8-flash-high at 1493 (a Flash model, non-qualifying). - [2026-09-17] Speculation intensified that a stealth Gemini 4 Pro checkpoint is being tested on Arena under the "gemini-3.8-flash" label: a model developers suspect is Google's unannounced Gemini 4 Pro appeared on the Arena benchmarking platform under the name "gemini-3.8-flash" on Sept. 17, with developer reports citing the internal codename Argon and leaked results indicating it outperformed GPT-6 Astra and Claude Fable 5.1 in coding, reasoning and agentic tasks . This remains unconfirmed by Google. - [2026-09-21] Fact-checking coverage stresses this is still unofficial: no new evidence has surfaced, and Google still hasn't commented on the Arena listing or confirmed anything about Gemini 4 Pro; Google has confirmed the training run, and Arena has historically been where Google tests early, but nothing else is confirmed—no name, no release date, no price, no benchmark . - [2026-09-21] Community/rumor timelines converge on October 2026 as the likely public release window for the next Pro-tier model (widely referred to as "Gemini 4 Pro"), though Google has not confirmed the model's identity or the leaked scores, and secondary coverage discusses October 2026 as a possible window without a firm public release date or API availability established . Other outlets (mindstudio.ai) note prediction markets giving roughly 68-85% odds of a Gemini 4 release by end of November 2026, underscoring genuine uncertainty on timing. - [2026-09-24 today] No qualifying "Pro"-labeled Gemini model has newly appeared on the Arena leaderboard since the market opened (May 18, 2026); the two most recent Gemini debuts on the leaderboard are Flash-tier (3.7 Flash, 3.8 Flash), which do not count under the resolution criteria. BOTTOM LINE: As of today, no new Gemini "Pro" model has been added to the Arena leaderboard since the market opened; Gemini 3.5 Pro appears to have been cancelled, and the next Pro-tier release is widely rumored (but unconfirmed by Google) to be "Gemini 4 Pro," possibly launching around October–December 2026 — meaning the outcome hinges entirely on a release that has not yet happened and whose eventual Arena debut score is unknown, though prior Gemini Pro debuts (3-pro at 1501, 3.1-pro at ~1500) have historically cleared 1490. NOT FOUND: No official Google announcement of a "Gemini 4" or "Gemini 3.5 Pro/4 Pro" model card, pricing, or confirmed release date; no confirmed Arena leaderboard listing under a "Pro" label newer than gemini-3.1-pro-preview; no verification that the "gemini-3.8-flash" stealth listing is genuinely a Gemini 4 Pro checkpoint rather than a true Flash update.

- [2025-03-25] Gemini 2.5 Pro's initial LMArena debut was called "the largest score jump ever" versus the prior leaders, and it "debuted at the top spot on the LMArena leaderboard 'by a significant margin.'" Gemini 2.5 Pro is now #1 on the Arena leaderboard - the largest score jump ever (+40 pts vs Grok-3/GPT-4.5) Its raw Elo at the time was well below today's inflated scale (leaderboard rescaling since then makes direct comparison to a 1490 threshold not meaningful for this data point). Source: rdworldonline.com. - [2025-11-18] The most recent unambiguous "Pro" debut prior to this market's period was Gemini 3 Pro, which debuted at the top of the LMArena text-reasoning leaderboard, posting a preliminary Elo score of 1501 based on pre-release community voting — the first LLM to ever cross the 1500 threshold, placing it above Grok-4.1-thinking (1484) and Gemini 2.5 Pro (1451) . That is well above the 1490 threshold. Source: venturebeat.com. - [2026-02-19/20] Gemini 3.1 Pro (a "Pro"-labeled point release) debuted very close to parity with the top model: as of its Feb 20, 2026 debut day, Arena's blind user voting presents a different picture... In the overall Text category, Gemini 3.1 Pro sits just 4 points behind the leading Opus 4.6. Contrary to the "dominant leader" narrative from official benchmarks, human blind testing shows the two models as essentially tied , implying a debut score in the high-1480s/1490s range. Source: smartscope.blog. - [2026-05-11] A later snapshot (months after debut, so not a pure debut score, but indicative of the model's plateau) shows Gemini 3.1 Pro sits at 1493 on Arena, sandwiched between Claude Opus 4.7 at 1503 and GPT-5.5 at 1484 . Source: helloai.com. - **Outside-view pattern**: Of the last three major Gemini "Pro" debuts (2.5 Pro, 3 Pro, 3.1 Pro), two clearly cleared what would now be an ~1490 bar (3 Pro at 1501; 3.1 Pro effectively ~1490-1496), while 2.5 Pro's debut was on an older, lower leaderboard scale not directly comparable. Google's Pro-tier models have consistently launched at or near the #1 position on Arena in the 2025-2026 period. - [2026-09-24] Polymarket itself already runs this exact market ("Next Google Gemini Pro Model: Arena Debut?"), pricing the "1490+" outcome at roughly 86% (85.9¢ "Yes"), versus 94-95% for "1480+", 73% for "1495+", 60% for "1500+", and 37% for "1510+". This is a direct market-based base rate/forecast for this specific event. Source: polymarket.com. - [2026-09-24] Market context notes accompanying the Polymarket listing state Google's Gemini 3.5 Pro (or a successor sometimes rumored as "Argon"/Gemini 4 Pro) remains delayed and in limited partner testing, with traders "watching for an official Arena debut or broader rollout before year-end." n/a (contextual note, not a citation-bearing claim). Source: polymarket.com. - [2026-08-23] Independent reporting confirms Gemini 3.5 Pro has repeatedly missed announced launch windows: Gemini 3.5 Pro is still unreleased as of 23 August 2026, more than three months after Google announced it at I/O on 19 May. It has now missed three targets (June, mid-July, early August), and has no model ID, no pricing and no launch date . Source: codersera.com. - [2026-08-01/02] Unverified leak reports claim brief, unconfirmed sightings of a "gemini-3.5-pro" identifier in Arena's blind-testing pool, but no official LM Arena notice, public leaderboard entry, or Google statement currently corroborates the specific gemini-3.5-pro Arena appearance... the claim that the unreleased model appeared on LM Arena is still supported by only one outlet's reporting and unverified community posts . Source: windowsforum.com. - [2026-08-05] Reporting links internal Google DeepMind leadership changes to the stalled Pro-tier roadmap: On 5 August 2026 Demis Hassabis stepped back to chairman of Google DeepMind and Alphabet chief scientist, with CTO Koray Kavukcuoglu taking day-to-day control... Jeff Dean left the same day. Fortune tied the reshuffle directly to stalled models and missed deadlines . Source: codersera.com. - [2026-06-26] An earlier Polymarket release-timing market ("Next Google Gemini Pro Model released on...?") shows persistent trader skepticism about near-term release, noting this delay, amid internal validation for advanced reasoning, 2-million-token context, and "Deep Think" capabilities, has traders assigning a modest edge to no release by July 31... Google's history of iterative previews before broad rollout, creates the tight clustering around July dates . Source: polymarket.com. BOTTOM LINE: The reference class of the last three "Pro"-labeled Gemini debuts shows Google's flagship Pro models have historically launched at or very near the top of the Arena leaderboard (Gemini 3 Pro at 1501; Gemini 3.1 Pro effectively ~1490s), an outside-view base rate favorable to clearing 1490 — and the live Polymarket market itself already prices this exact outcome at roughly 85-86% "Yes," reflecting trader confidence tempered by uncertainty over whether the delayed Gemini 3.5 Pro (or a successor) will even debut by year-end and how it will score amid a competitive field near the 1500 mark. NOT FOUND: An authoritative, single confirmed numeric "debut score" for Gemini 3.1 Pro's very first appearance on the LMArena text leaderboard (sources give a range of ~1490-1496 from slightly different snapshots); confirmation of whether Gemini 3.5 Pro or another new "Pro" model has yet formally appeared on the Arena leaderboard as of September 24, 2026.

- [2026-09-17] Multiple independent developer reports (via X posts) claim a stealth checkpoint appeared on Arena under the label "gemini-3.8-flash" that outperformed the genuinely-released Gemini 3.8 Flash on coding, SVG generation and agentic tasks, with leaked benchmarks showing the model carries the internal codename Argon and has not been officially announced by Google, but leaked benchmark results shared by developers suggest it outperformed OpenAI's GPT-6 Astra and Anthropic's Claude Fable 5.1 in coding, reasoning and agentic tasks , source dataconomy.com. - [2026-09-14] One source dates the "Argon" codename claim precisely: "Is Gemini 4's codename argon? A 14 September 2026 post on X claimed so. Google has not confirmed it." (projedefteri.com), underscoring that even the leading YES-supporting narrative traces to unverified social posts, not Google. - [2026-09-22] A tester's community estimate specifically pins a launch window: "A tester now places a possible public launch 'around the middle of next month,' which would mean roughly mid-October, but this remains a community estimate rather than a Google schedule (SPAC89)." (kie.ai), i.e., a debut inside the market's trading window (closes 2027-01-01) is plausible per rumor. - [2026-09-21] Officechai's reporting frames the strategic mechanism for YES: Google appears to have abandoned the weaker "Gemini 3.5 Pro" bridge model entirely in favor of shipping a stronger flagship directly, noting "There had been some speculation that Gemini 3.5 Pro was being tested out on Arena at that time, but it appears that the results didn't fully satisfy Google... But the company has put its weight behind Gemini 4 Pro, and it appears that Gemini 3.5 Pro might end up getting skipped altogether." This implies the next Pro-labeled debut will be a more capable, higher-scoring model than a modest point-release would have been. - [2026-09-21] Leaked pricing details suggest an aggressive, competitive positioning consistent with a genuine flagship push rather than a incremental patch: "Circulating information indicates pricing of $2.25 per million input tokens and $11.25 per million output tokens, with a possible 10-million-token input limit and cross-conversation permanent memory." (biggo.com), a spec set that if real would target frontier-tier quality, not a Flash-tier score band. - [Historical mechanism] The market itself distinguishes score thresholds finely (1480/1490/1495/1500/1510), and per the live Polymarket-derived tracker fetched today, "Markets price the chance that the next Gemini Pro model added to the Arena.AI leaderboard debuts at a score of at least 1480 at 94.1%" while the 1490 line specifically sits at 91.3% with a same-day jump of "+12.4" points — a sharp repricing toward YES coincident with the September 17-21 Arena stealth-model reports, i.e., the market itself reacted bullishly to the "gemini-3.8-flash"/Argon leak as evidence the next debut will be strong. - [Current leaderboard baseline] As of the most recent fetched snapshot, competitive Arena text scores cluster tightly near or above 1490-1500 (e.g., Claude Fable 5 ~1505.7, Fable 5.1 ~1498.5, per felloai.com's September update), meaning any genuinely frontier-grade Gemini successor competing for the top spot would very likely need a score in that same 1490-1510+ band to be considered "frontier" at all — consistent with prior Gemini Pro debuts (3 Pro at 1501; 3.1 Pro near 1490-1496) clearing 1490. - [Mechanism for YES independent of "Gemini 4" branding] The resolution criteria explicitly count any Pro-labeled release, including a possible interim "gemini-3.1-pro" refresh or renamed "gemini-4-pro-preview" checkpoint; because Google DeepMind already confirmed on July 21 that pre-training for "Gemini 4" was underway (per the July 21 post cited in prior research), and internal checkpoints reportedly exist now (per the nokiapoweruser September 2026 "first checkpoint" report cited above), the technical prerequisite for a debut (a trained, testable Pro-tier model) already plausibly exists — leaving public/Arena listing as the main remaining gating step rather than model readiness. BOTTOM LINE: The strongest case for YES rests on (a) strong circumstantial evidence — echoed across at least six independent outlets — that a genuine next-generation Pro-tier Gemini checkpoint ("Argon"/Gemini 4 Pro) is already being tested on Arena disguised as "gemini-3.8-flash," reportedly beating GPT-6 Astra and Claude Fable 5.1 on coding/agentic benchmarks, which would imply a frontier-caliber, likely 1490+ debut score consistent with the last two genuine "Pro" launches (3 Pro at 1501; 3.1 Pro ~1490-1496); and (b) the live Polymarket-adjacent tracker itself has already re-priced the 1490+ outcome up to roughly 91% following these September leaks, reflecting trader belief that if/when a Pro model debuts, it will very likely clear 1490. NOT FOUND: No official Google confirmation that the "gemini-3.8-flash" Arena listing is genuinely a Pro-tier or "Gemini 4 Pro" checkpoint (Google has not commented); no confirmed model card, Arena listing under a "Pro" label, or numeric score for any post-3.1-Pro Gemini model as of September 24, 2026; no firm, Google-sourced release date (October 2026 remains a community/X-sourced estimate only).

- [2026-09-24] The live Polymarket "Next Google Gemini Pro Model: Arena Debut?" contract itself prices real, non-trivial NO risk at every threshold: the current frontrunner for "Next Google Gemini Pro Model: Arena Debut?" is "1480+" at 95%, meaning the market assigns a 95% chance to that outcome. The next closest outcome is "1490+" at 86% — i.e., traders themselves put roughly 14% on a sub-1490 debut, and the probability decays fast at higher thresholds (73% at 1495+, 60-62% at 1500+, ~38% at 1510+), reflecting genuine two-sided uncertainty rather than near-certainty of a clean "Yes." (polymarket.com) - [2026-09-24] A companion Polymarket market on timing — "Next Google Gemini Pro Model released by...?" — shows meaningful chance of a release-and-debut delay past the near term: the current frontrunner is "October 31" at 78%, meaning the market assigns a 78% chance to that outcome. The next closest outcome is "October 23" at 71% . That leaves roughly a one-in-five (or higher) chance the model isn't even out by Halloween, compressing the runway before the market's hard Dec 31, 2026 deadline and raising the odds of a rushed, less-polished debut or of missing the window entirely. - [2026-09-24] Sibling Polymarket contracts about Google's competitive standing cut against a triumphant debut narrative: one related market shows "Will Google have a #1 AI model by December 31, 2026?" at 27% , implying traders think it's more likely than not that Google's next flagship will *not* top the field even after launch — consistent with a debut that lands mid-pack near 1490 rather than decisively above it. - [2026-09-21] Independent tracking of the September 17 "gemini-3.8-flash" stealth-listing saga confirms the story keeps accumulating rumor-volume without new hard evidence: as of the September 21 update, more outlets, including TechBriefly and Dataconomy, are now covering the same September 17 Arena leak, still citing the same unverified benchmark table and the same Qwinah screenshot. No new evidence has surfaced, and Google still hasn't commented on the Arena listing or confirmed anything about Gemini 4 Pro. The story keeps growing. The confirmed record hasn't moved an inch . This is a critical resolution-criteria risk: if Google's true next-gen Pro-tier checkpoint keeps testing under a Flash-labeled or anonymous name (as it reportedly has since Sept 17), the market may never see a qualifying "Pro"-labeled Arena entry before the deadline. - [2026-09-21] The same tracker notes the September benchmark leak is internally inconsistent and contested: this benchmark table does not fully match the suspected backend screenshot "dug out" by Qwinah; the score of Astra as a reference item in the benchmark table also does not conform to the official data; neither of the two materials has official endorsement from Google, Arena or relevant benchmark institutions , and one Reddit commenter called the leaked benchmark table "fabricated" (getneotiler.com, Sept 19). Fabricated or exaggerated leaks undercut confidence that any real debut score will clear a specific numeric bar. - [2026-09-19] There is direct historical precedent for stealth listings that never get formally claimed or resolved as a distinct model, which would leave this market unresolved on the merits: in mid-August, an anonymous model called Ox Alpha surged into the top three on the leaderboard within 48 hours, and no one has claimed it to this day... what is truly rare is not the anonymity itself, but reusing the name of an already-released in-house model . If Google repeats this pattern with its true next Pro-tier model, no compliant "Pro"-labeled debut may ever appear. - [2026-08-24] Independent, market-based Gemini-4 timing analysis converges on Q4 2026 at earliest, with real risk of slipping toward year-end: Google has publicly confirmed that Gemini 4's pre-training run is underway, and prediction markets currently give it roughly an 85% chance of shipping by November 30 , while independent analysts generally expect a November or December launch, which would put Gemini 4 roughly a year after Gemini 2.0's debut and would give Google a full pre-training and safety evaluation cycle after its July confirmation (mindstudio.ai). A November/December landing leaves little buffer before the Dec 31 deadline and the built-in 7-day leaderboard-unavailability grace period, raising the chance of an automatic "No" via deadline arithmetic alone. - [2026-09-19] Independent reporting on release-date consensus flags the same "no confirmed date" problem repeatedly cited by colleagues, but adds a fresh, more granular split: some accounts on X point to October, saying an internal checkpoint is already live in Arena. 9to5Google (July 26) and TestingCatalog (August 24) both estimate November or December, based on Google's past release cadence and preparation traces found in the desktop app — i.e., the two most-cited independent trackers explicitly favor a later (November/December) window over the more optimistic October rumors, tilting toward tighter deadline risk. - [2026-09-19] Google's own internal quality bar appears to have risen sharply, per the same source's account of the scrapped 3.5 Pro: the earlier candidate was reportedly abandoned because internal successors "weren't sufficiently better than the Flash series," per the Wall Street Journal (via techbriefly.com's Sept 21 summary) — suggesting Google is now holding out for a model good enough to clear a much higher competitive bar, which raises variance: either a genuinely dominant debut (favoring Yes) or continued delay/further cancellation risk (favoring No via deadline). - [2026-09-24] Trading volume and market structure show this is a well-monitored, liquid market ($185,954 traded on the arena-debut contract as of today) with active price discovery across five distinct thresholds, meaning the current ~86% "Yes" on 1490+ already reflects informed aggregation of all the delay/stealth/quality risks above — not a naive extrapolation from past Gemini debuts alone. (polymarket.com) BOTTOM LINE: The strongest NO case rests less on "Google's Pro models usually score low" (they haven't) and more on process/timing risk: Gemini 3.5 Pro was scrapped, the true successor ("Gemini 4 Pro"/"Argon") remains unconfirmed and reportedly still testing under non-Pro labels (Flash-branded or anonymous) as of late September, independent trackers lean toward a November-December release window that leaves thin buffer before the Dec 31 deadline, and Polymarket's own market prices roughly 14% NO risk on the 1490 threshold even conditional on a qualifying debut occurring at all — with additional unpriced risk that no compliantly-labeled "Pro" model appears before the deadline at all, which would resolve NO regardless of any underlying model's quality. NOT FOUND: No official Google confirmation of a "Gemini 4 Pro" name, model card, or release date; no verified (non-leaked) Arena leaderboard entry labeled "Pro" newer than gemini-3.1-pro-preview; no authoritative record of Arena's historical rate of leaving anonymous/mislabeled stealth models permanently unclaimed (only the single "Ox Alpha" precedent found) to quantify the "never properly labeled" tail risk.